Audit professionals’ perceptions of
CEO narcissism and audit committee
strength and the influence of auditor
expertise and narcissism at a risk
assessment
Abstract
The purpose of this study is to investigate the perceptions of audit professionals (certified and non-certified) of CEO narcissism and audit committee strength at a risk assessment experiment. Moreover, the purpose of this study is to explore the effects of auditor narcissism and expertise on this risk assessment at an audit. Previous studies have shown the increasing effects of client CEO narcissism and the decreasing effect of the client’s audit committee on the perceived audit risk. The final regression model with all variables that have been studied in this experimental study included indicates that CEO narcissism, audit committee strength, and the interaction effect between these variables have an insignificant effect on the perceived audit risk. Furthermore, the results indicate that auditor expertise has a significant effect (two tailed p-value <0.05) on the perceived audit risk, where auditors with a higher level of expertise reported a higher perceived audit risk. Finally, the results of this study indicate that auditor experience is a surrogate for auditor expertise. No evidence was found for potential effects of auditor narcissism on the perceived audit risk at a risk assessment.
Master’s thesis MSc Accountancy & Controlling Douwe Bakker s2081032
12-1-2018
University of Groningen, Faculty of Business Faculty of Economics and Business Supervisors: J. Emanuels and M. L. Lu
1
2
1 Introduction
In recent times, multiple substantial reporting-related management scandals have occurred. Cases such as those of Enron and WorldCom have led to distrust concerning the integrity of firms in the United States (Carson, 2003). To regain confidence in the financial market, the Sarbanes-Oxley (SOX) Act was created. This act requires public companies and their external auditors to assess the effectiveness of their internal control system related to their financial reporting; it contains provisions to make the management of the company more accountable, among other legislation. However, regardless of the improvements in the external audit standards since the introduction of the SOX Act, the problems concerning fraudulent financial reporting have persisted (Hogan et al. 2008).
Since the stock market crash in the early 21st century, now more than ever the ethics of managers
and executives are being called into question (Carpenter & Reimers, 2005). Financial reporting scandals that question management ethics have occurred in the Netherlands as well, such as the controversial case of Ahold in 2002: the company included revenue in its consolidated statements that should not have been recognized. In addition, around 2012, the Dutch listed company Imtech covered up the losses of its Dutch subsidiary by reporting its German revenue.
A Dutch survey found that 70% of the top managers placed themselves in the 25% best managers (Al, 2010). This overconfidence is related to financial reporting concerns (Schrand & Zechman, 2012). Moreover, Malmendier and Tate (2005 and 2008) found that there is an association between overconfidence and distorted financial decisions. When a CEO tries to inflate his self-image by manipulating the firm’s financial performance, this is often related to that manager’s narcissism (Olsen, 2004).
This personality trait, narcissism, is widespread among CEOs (Peterson, Galvin & Lange, 2012). Some amount of self-confidence can be beneficial, however narcissistic behavior is related to negative consequences such as willing to cross the line to commit fraud for their admiration and grandiosity (Rijsenbilt & Commandeur, 2013). Auditors have to address these risks their audit, with regards to management tendencies in their audit
Client-specific risks are identified during an initial risk assessment at the planning phase of an audit. Performing a risk assessment and the decision-making behind it is a highly complex process. Recently, the Dutch Association for Registered Accounting Professionals (NBA) identified a gap regarding fraud risk factors in the education of Dutch certified public accountants (CPAs) and introduced an obligatory course for all these professionals in 2017 (Constansia & Mulder, 2017). An important factor in fraud risks is management: to truly understand these specific risks, the auditor has to understand the behavior and underlying motivations of the management. Important here is that risks are often not quantifiable and highly subjective. What is the difference, for example, between a manager who is slightly overconfident and one who is highly narcissistic? Unlike quantitative key performance indicators such as profit, these behavioral traits are more difficult to identify. Moreover, how an auditor assesses such behavioral traits could impact the further audit.
An auditor can mitigate these risk by performing additional audit procedures after the identification a specific risk. However, mitigating factors may also exist within the company with regard to these management-specific risks. The present study takes into account the mitigating effect of corporate governance in a two-tier setup. From an agency theory perspective, these independent agents
3 are hired as supervisors of the management; they are primarily concerned with their perceived value on the job market and have no incentive to report higher profits. In contrast, managers usually have a variable component in their compensation linked to the performance of the firm. This study specifically focuses on one body of the supervisory board, the audit committee (AC). Research shows a negative relationship between a strong AC and the occurrence of EM (Klein, 2002 and Van der Zahn and Tower (2004). Hence, management is less likely to influence earnings when there is a strong AC.
In this thesis, the effects of CEO narcissism and a strong AC on the auditor's perceived audit risk are investigated using an experimental setup. All other variables were held constant to study the effect of these two factors. The aim is to answer the following research question:
Does having a narcissistic CEO affect an auditor’s risk assessment at an audit, and does this
relationship change with the countervailing power of an independent and strong audit committee?
Furthermore, this study examines the decision-making process of an auditor with regard to the risk assessment and will investigate if certain auditor characteristics impact the risk assessment. Little research has investigated the effect of auditor characteristics on complex audit decisions such as risk assessments. Thus, this study provides new insight in this regard. Moreover, this research method provides the unique opportunity to study how certain personal factors affect auditors’ decision-making process in a risk assessment where CEO narcissism and AC strength vary among the experimental treatments.
Besides the main research question, two auditor characteristics that could have an effect on the decision making process of are also investigated in this present study. These two characteristics are: expertise and narcissism. Expertise is the ability acquired by practice to perform qualitatively well (French and Steinberg, 1989). In this study, it is hypothesized that auditors with higher expertise are better at performing risk assessments.
Besides expertise, this study also examines auditor narcissism. Research shows that narcissists have distorted judgements of their own capabilities and risk-based decision-making (Twenge et al. 2008). Hence, this study explores whether the narcissistic trait of an auditor negatively affects the decision making in the risk assessment process.
These two co-variables lead to the following two sub-questions:
How does auditor expertise affect the auditor’s risk assessment at an audit? How does auditor narcissism affect the auditor’s risk assessment at an audit?
This study is rooted in the agency theory literature and further investigates the relationship between different agents in an experimental setup: the external auditor, the audit committee, and a CEO hired by the principals to manage their capital. Moreover, this study embodies the principles of Hambrick and Mason’s upper echelon theory.
In this study, a classical two-by-two double-blind experimental design was used. The subjects, a group of professional Dutch auditors from different Big Four firms assessed the perceived audit risk in the planning phase of an audit given a CEO high/low in narcissism and a strong/weak audit committee.
This study contributes to literature further evidence to the literature regarding the effects of CEO narcissism on the perceived risk by the (external) auditor. Johnson et al. (2013) and Judd et al.
4 (2016) already investigated this relationship in an experimental and archival study. However, to the best of the author’s knowledge, the present study is the first to investigate the countervailing power of an audit committee in a large experimental set up. This thesis introduces the effect of the audit committee on the perceived audit risk given the manager’s level of narcissism. Moreover, it contributes more insight regarding the effect that certain auditor characteristics, namely auditor expertise and auditor narcissism, have on the risk assessment of a client at an audit.
This thesis also provides practical insights regarding the risk assessment of management fraud indicators by professional Dutch auditors. In 2008, Hoogan et al. wrote that little guidance was available in the auditing literature regarding indicators of fraud (Hogan et al. 2008). Recently, the Dutch Association for Auditors (NBA) also acknowledged this gap in the Netherlands, and in 2017 introduced an obligatory course for certified public auditors on fraud risk. However, the underlying problem is that the current curriculum of public auditors in the Netherlands does not include courses on organizational behavior (R. Constansia & Mulder, 2017). The present study provides insight into how Dutch professional auditors assess the risks regarding a narcissistic/non-narcissistic CEO and a strong/weak AC. The Dutch NBA can use the results of this study as a starting point to determine whether more organizational behavior courses are required in the curriculum for certified public auditors. Or that in courses or permanent education for the certified auditor additional attention has to be given to the audit committee with regards to the internal environment of the company at an audit.
These previous mentioned aspect can also being taken in to consideration by accounting firms for their internal education. Moreover, for these firms this study provides information on how auditor expertise and narcissism affects risk assessments by auditors.
This thesis is organized as follows. Chapter 2 introduces the scientific literature relevant to this research and presents the hypotheses. Subsequently, Chapter 3 elaborates the methodology of this experimental study, and Chapter 4 analyzes the results. Finally, in Chapter 5 a conclusion is formed and the findings of this study are discussed in the perspective of current literature.
5
2 Literature Review
This chapter discusses the underlying literature used in the present study. It first presents the main theory in which this research is rooted, and then defines the important concepts. Finally, the hypotheses are developed and visualized.
2.1 Economic Theories
As introduced in Chapter 1, this thesis is rooted in the agency theory. In addition, the concepts of the upper echelon theory are also used. Hence, this section first introduces the agency theory and explains why it is applicable to this study. Next, the section argues for the use of the upper echelon theory.
2.1.1 Agency Theory
Classical agency theory investigates the principal-agent relationship, where it is assumed that managers (agents) act in self-interest instead of the principals’ interests, and the shareholders (principals) want to avoid that risk aligning interests by contracting (Jensen & Meckling 1976, Eisenhardt, 1989). The assumed self-interest of managers leads to information asymmetry and moral hazard for the shareholders if this risk is not mitigated. The agency theory assumes the human factors of self-interest, bounded rationality, and risk aversion. That makes this theory perfect for the present study.
The agency theory presumes that the manager (agent) has an incentive to manipulate results to increase his remuneration. He does not have the risks that the AC and external auditor have. On the other hand, the external auditor (agent) wants to give accurate assurance about the client's financial statements to the shareholders and society (principals) to avoid potential lawsuits. To this end, the auditor will conduct an initial risk assessment and, based on that assessment, perform procedures to reduce the audit risk to an acceptable level. Furthermore, independent board members such as the AC want to protect their own image on the job market and want to reduce the risks of inaccurate financial statements, as this would devalue their position (Boo and Sharma, 2008).
This study provides further insight from an external auditor perspective into the relationship between shareholders and society (principals), managers, and the audit committee (other agents).
2.1.2. Upper Echelon Theory
This thesis also employs the upper echelon theory. This theory states “that organizational outcomes, the strategic choices, and performance levels are partially predicted by managerial background characteristics” (Hambrick and Mason, 1984). An important concept in the present study is the effect of key actors in the organization on its outcomes. This study focuses on the lead auditor as a key actor who could impact the audit engagement. The key actors of the firm that is being audited are the CEO and the AC. Hence, the upper echelon theory is highly applicable. It is hypothesized in this study that the CEO can have an impact on the organizational outcomes of his/her company that could in turn impact the audit engagement, leading to a higher audit risk. This can either be due to manipulation of financial performance or other factors such as the increased likelihood of errors. Finally, it is presumed that a strong AC has the ability to mitigate the CEO’s potential effect on the audit risk. In the next part the main concepts of the theoretical framework of my study will be outlined and the underlying arguments for the hypotheses will be introduced.
6
2.2 CEO Narcissism
2.2.1 CEO Influence
The executive board and especially the CEO are in a major position to influence the results and strategy of the company (Finkelstein and Hambrick, 1996). The present study focuses on the CEO, in line with the upper echelon theory. An argument for this is that the tone set at the top by the CEO has an impact on others in the organization. In this vein, current research suggests that CEOs can impact the financial reporting decisions made by other employees (D’Aquila & Bean (2011).
Managers can have multiple arguments to manipulate the financial performance of their firm. The most important reason for them to do so is because their bonus is linked to this performance. The literature defines this phenomenon of managers manipulating earnings as earnings management (EM). Research even shows that the CEO’s tone can impact the financial reporting decisions made by other employees in the organization (D’Aquila & Bean (2011). Hence, the CEO has a highly influential position in the company to manipulate earnings. When a CEO tries to inflate his self-image by manipulating the firm’s financial performance, this is often related to that manager’s narcissism (Olsen, 2004).
2.2.2 Narcissism
The following discusses the trait of narcissism. Academics such as de Vries and Miller (1985) distinguish between reactive, self-deceptive, and constructive narcissism in a managerial context. However, in the present study no such distinction is made. Instead, narcissism is defined as one theoretical construct.
Narcissism has its roots Greek mythology. A man called Narcissus fell in love with his own reflection in a pool. He could not leave the beauty of his reflection, which eventually led to his death. Later, Freud (1914) proposed the concept of narcissism as a personality disorder, and Raskin and Hall (1979) introduced clinical narcissism in the psychology literature (Paulhus & Williams, 2002).
The American Psychiatric Association (2000) defines the narcissistic personality disorder as a “pervasive pattern of grandiosity” combined with a “need for admiration and lack of empathy.” Moreover, narcissistic people overestimate their IQ (Gabriel, Critelli, & Ee, 1994) and their own contribution to a collaboration relative to ratings by group member and independent observers (Gosling, John, Craik, & Robins, 1998). Furthermore, narcissistic people look for scenarios that offer possibilities to demonstrate their grandiosity (Morf, Weir, and Davidof 2000; Morf & Rhodewalt, 2001). Morf and Rhodewalt (2001) describe that narcissistic people are “living on an interpersonal stage with exhibitionistic behavior and demands for attention and admiration but respond to threats to self-esteem with feelings of rage, defiance, shame, and humiliation.” These authors further state that if a narcissist has to deal with personal failure, he or she tries to find options to undo this personal failure.
In conclusion, narcissistic behavior has been widely researched and is a unique behavioral trait that is even labeled as a personality disorder in clinical research. The present study uses the definition proposed by the American Psychiatric Association.
7
2.3 Auditor’s Perception of Audit Risk
To provide assurance about the financial or non-financial statements of company, an auditor conducts an audit. Presently, financial statement audits use a risk-based approach, where a risk assessment of the company is initially performed. This risk assessment serves to indicate in what processes and accounts a risk of material misstatement (ROMM) can occur. Given that there is no unlimited budget for an audit, procedures are performed to address the risk found in the initial risk assessment to reduce that risk to an acceptable level.
Much research has examined how auditors react to this perceived audit risk. Audit risk can be defined as the risk that the accounting firm will suffer a loss on the audit engagement through litigation or through a lack of profitability of the engagement (Johnstone, 2000). To prevent a loss by for example litigation the client integrity is being assessed by the auditor. Hernandez and Groot (2006) found that audit partners react to lower levels of client integrity with additional fraud risk assessments during the initial phase of an audit. This is in line with the findings of Johnson, Kuhn, Barbara, Aposolou, and Hassel (2012).
The present study focus on the audit risk, and not only the fraud risk. For instance, Olsen et al. (2014) show that narcissistic CEOs are more focused on short-term results, as they invest more in activities that improve short-term earnings despite the possible long-term negative risks. Therefore, the narcissistic tendencies of a CEO can also lead to higher non-fraud-related audit risks.
Hypothesis Development
The personality trait of narcissism is widespread among CEOs (Peterson, Galvin & Lange, 2012). A moderate amount of narcissism can be beneficial, as narcissists possess the charisma and grand visions that are vital to leadership (Rosenthal & Pittinsky, 2006). For instance, Olsen, Dworkis, and Young (2013) found a significant relationship between narcissism among CEOs and earnings per share and stock valuation. Hence, moderate amounts of narcissism can have a positive effect for shareholders.
On the other hand, research shows that CEO narcissism also has disadvantages. A recent study that included all S&P 500 CEOs with a tenure of longer than 3 years found that CEO narcissism is significantly positively related to managerial fraud (Rijsenbilt & Commandeur, 2013). In addition, the study confirms the psychological perspective that narcissistic CEOs are willing to cross the line for their admiration and grandiosity. Auditors have to address this risk in their audit. Presently, researchers agree that narcissistic CEOs pose an increased audit risk (Judd, Olsen, and Stekelberg, 2016; Johnson et al. (2012) among others).
Zu and Chen (2014) found that CEOs prefer new directors who are similar in narcissistic tendency or have worked for other narcissistic CEOs. This has the effect that a narcissistic “tone at the top” by the CEO can lead to an overall more narcissistic board, which in turn leads to more audit risk. This is in line with the upper echelon theory introduced in the first part of Chapter 2.
Furthermore, Kuhn, Barbara, Aposolou, and Hassel (2012) show that narcissistic client behavior and fraud motivation are significantly and positively related to auditors’ risk assessment. This is not in line with the findings of Hogan et al. (2008), however, whose results indicate that auditors do not make significant adjustments to audit plans as a result of higher fraud risk assessments. On the other hand, Kirzrian, Mayhew, and Sneathen (2005) suggest that when auditors notice a lack of integrity during the planning phase of an audit, they use more external audit evidence to
8 mitigate this integrity risk. Hence, the initial presumed audit risk is still taken into account by obtaining more external audit evidence, even though the audit plans themselves are not changed. The relationship between narcissistic client behavior and the auditor's risk assessment is also confirmed by Johnson et al. (2012). They used an experimental setup to show this positive relationship. Johnson et al. focused specifically on the perceived fraud risk, however, whereas the present study focuses on the audit risk as a whole. An argument for this is that Olsen et al. (2013) found that narcissistic CEOs are more focused on short-term results, as they invest more in activities that improve short-term earnings despite the possible long-term negative risks. In this vein, it is argued here that narcissistic CEO behavior does not only lead to increased fraud risk but also to more risk regarding the CEO’s focus on short-term earnings. This short-term opportunism is contradictory to the conservative perceptions of accounting principles on which the auditors base their judgement. Hence, this short-term focus of the CEO will lead to more audit risk in general. Hence, the following hypothesis is formulated:
Hypothesis 1: Highly narcissistic tendencies of a CEO will lead to a higher perceived audit risk by the auditor at a risk assessment compared to a CEO with a low level of narcissistic tendencies.
2.4 Audit Committee
First, the supervisory board in general is elaborated. Moreover, the function and purpose of the Audit Committee (AC) within the supervisory board is explained based on best practice guidance’s from the Corporate governance code.
An important countervailing power to the executive management is the independent supervisory board. Fama and Jensen (1983) describe the function of this board from an agency theory perspective: it is a structure that monitors that the management of the company, in such a way that the management makes decisions according to the preferences of the shareholders. According to Boo and Sharma (2008), independent board members are better at monitoring the company because they have no financial interest other than enhancing their own image on the job market. In contrast, managers are remunerated based on their performance and they do have an incentive to manipulate earnings, as discussed in at this first part of Chapter 2.
In the Netherlands, a two-tier board structure is prevalent, where the CEO is not the chairman of the supervisory committee. Since January 1st 2012, it is possible to implement a one-tier board
structure, but most companies continue to employ the two-tier approach.
Recent Dutch corporate governance reforms have advocated the use of independent sub-committees of the board of directors (De corporate governance code 2003, 2008, and 2016). This code is a best practice, whereby if companies do not conform to the code, companies have to explain why in their annual report. One of the principles is that if a supervisory board consists of more than four members, it has to appoint an audit committee, selection committee, and remuneration committee. In the Netherlands, the supervisory board often gives specific tasks to the audit committee (Maassen, van de Bosch & Volberda (2005).
In order to give an understanding of the role of the audit committee there is elaborated on the tasks of the Audit Committee. The Dutch corporate governance code outlines the following responsibilities for audit committees (Monitoring committee corporate governance code, 2016):
9 a) the operation of the internal risk management and control systems;
b) the provision of financial information by the company, such as the choice of accounting policies, application and assessment of the effects of new rules, and information about the handling of estimated items in the financial statements;
c) compliance with recommendations and observations of internal and external auditors; d) the role and functioning of the internal audit function;
e) the company’s policy on tax planning;
f) relations with the external auditor including, in particular, his independence, remuneration, and any non-audit services for the company;
g) the financing of the company; and
h) the applications of information and communication technology.
Now that the concept of AC has been explained, the following section develops the related two hypotheses.
Hypothesis development
In recent research, a strong AC is often being measured by quantitative proxies for a strong AC such as the number of AC meetings which is a proxy for AC activity and for example the number of AC members included with a financial background as a proxy for financial expertise. These proxies have been widely used in quantitative research to provide generalizable results. Therefore, this argumentation also focuses on often used indicators of AC strength: AC independence, AC financial expertise, and AC activity.
As previously mentioned, Boo et al. (2008) theorize that supervisory board members have no direct financial interest in the company and are only directly concerned with their own image on the job market. Hence, it is hypothesized here that strong and independent ACs specifically want the external auditor to provide a high-quality audit to avoid negative audit opinions, financial restatements, and statements about internal control weakness, thereby allowing them to maintain their image in the labor market.
Current research shows that a strong AC is related to an increased audit fee (Abbott et al. 2003, and Goodwin-Stewart & Kent, 2006). In line with Abbott et al. (2003), it is assumed here that higher audit fees lead to a higher level of assurance as more budget is available to perform the required procedures, and that this supports the oversight role of the management-auditor relationship. From an AC perspective, this higher fee leads to more assurance by the external auditor and a lower likelihood of reputational damage through restatements or internal control weaknesses. This is in line with the goal of the external auditor. That is to mitigate the audit risk to such a low level that a reasonable amount of assurance can be given with regards to the financial statements. If the auditor identifies the AC as strong and independent, thus not dependent of the management this should lead to a lower perceived audit risk. As an internal department of the organization is monitoring and reducing the same risks that the auditor tries to reduce at his audit by performing audit procedures.
Besides it is for an auditor important to assess if the auditor has the ability to reduce these audit relate risks. Therefore the auditor has to analyze the strength of the audit committee in order for the AC to have an audit risk reducing effect. There has been plenty of research with regard to AC strength.
Xie, Davidson, and DaDalt (2003) suggest that ACs’ activities and their financial expertise are important factors in their capability to act as financial monitors. This is in line with other studies. For instance, a negative association has been found between AC independence, AC activity, AC
10 financial expertise, and the occurrence of a financial restatement (Abbott, 2004). Furthermore, Krishnan (2005) shows that a company with an independent AC and financial expertise in an AC is less likely to be associated with internal control problems, which have been directly associated with fraud (Bell and Carcello 2000). Furthermore, there is a negative relationship between AC independence and the occurrence of EM (Klein, 2002 and Van der Zahn and Tower (2004). We conclude that potential indicators for a strong AC are activity, financial expertise and independence.
It is concluded that goals of the AC align with the goals of an external auditor. Hence, an AC with strong characteristics such as financial expertise, independence that is active is thought to lead to a decreased perceived audit risk by the auditor:
Hypothesis 2a: A strong AC leads to a lower perceived audit risk by the auditor at an risk assessment compared to a weak AC.
Carcello et al. (2011) indicate that the involvement of the CEO in the hiring process of the AC has a negative impact on the effectiveness of the AC. Moreover, the results of their study suggest that the effectiveness of an AC with strong characteristics such as financial expertise and AC activity is decreased when the CEO has a direct influence on the hiring process of the AC.
Zu and Chen (2014) found that CEOs prefer new directors who are similar in narcissistic tendency or have worked for other narcissistic CEOs. The effect of manager narcissism on audit risk was explained in Chapter 2.3. This preference of the CEO leads to a more narcissistic tone at the top set by the CEO, which can lead to an overall more narcissistic board and AC. In turn, this leads to a higher audit risk. However, it is predicted here that this will not be the case if there is a strong and independent AC, where the CEO has no direct influence on the nominating process.
Moreover, research shows that indicators of a strong AC are related to lower levels of earnings management (Klein, 2002 and Xie et al. (2003). Hence, the CEO is less likely to influence the earnings of his company if the company has a strong AC. Besides, the study of Abbott, Park and Parker (2000) suggests that the presence of an active and independent AC, which in this study have been used as indicators of a strong AC, are associated with a lower occurrence of fraud and non-fraudulent misstatement. Therefore, it is expected that a CEO with highly narcissistic tendencies has a lower opportunity to commit fraud and earnings management if a strong AC is in place. This lower opportunity for earnings management and fraud for a highly narcissistic CEO due to a strong AC should lead to a lower perceived audit risk by the auditor at a risk assessment. This leads to the following hypothesis:
Hypothesis 2b: A strong audit committee has a negative (moderating) effect on the perceived audit risk by the auditor at a risk assessment given the highly narcissistic tendencies of the CEO.
2.5 Personality Characteristics of Auditors
Besides the above hypotheses, this study also investigates whether auditor expertise and narcissistic tendencies influence the perceived risk by the auditor. This section first discusses auditor expertise, followed by auditor narcissism. Then, hypotheses are formulated.
2.5.1 Auditor Expertise
The present study investigates whether auditor expertise affects the perceived risk by the auditor, and specifically whether it has an impact on the risk identification concerning CEO narcissism.
11 In line with French and Steinberg (1989), expertise is defined in this study as “the ability acquired by practice to perform qualitatively well in a specific domain,” as this is the most widely accepted operational definition. Furthermore, these authors state that more practice leads to better performance with diminishing effects as expertise increases.
In this thesis, it is argued that there is a difference between auditor expertise and auditor experience. Beck and Solomon (1989) state that experience is unlikely to be a surrogate for expertise and expert judgement/decision performance in many public accounting tasks typically examined by accounting researchers. This is in line with Ashton (1991), whose results suggest that auditors do not gain expertise with regard to errors though experience, as experience with errors is rather limited.
Boner and Lewis (1990) also distinguish between task-specific knowledge (expertise) and experience. They found that task-specific knowledge assisted the performance of experienced auditors in both selection and weighting of the risk in an analytical risk assessment. This is supported by Knapp and Knapp (2001). In the present study, subjects were asked to perform a specific task: a risk assessment. Hence, expertise was used in the study.
Hypothesis development
It is presumed that auditor expertise has a substantial impact on the auditor’s risk assessment and thereby his or her perceived audit risk. Current research on the subject usually covers a specific aspect of expertise, such as general experience or specific industry expertise, as it is difficult to measure expertise in large validated quantitative set ups for a specific task. Therefore, only theory is available, and not much evidence regarding the whole concept of auditor expertise. Hence, the hypothesis development considers several aspects of auditor expertise.
Krishnan (2003) and DeBoskey and Yang (2012) found that clients of non-specialist auditors report higher discretionary accruals that indicate that more earnings management has been done by the management of the company. These results show that in a quantitative study, a specific aspect of expertise, industry expertise, influences the occurrence of earnings management which is a risk for the auditor that can affect the overall audit quality.
Besides, the experimental qualitative study of Shelton (1999) found that experienced auditors are better at processing irrelevant information regarding going concern judgements. This is highly applicable to a risk assessment for an auditor. To perform an accurate risk assessment, it is essential for auditors to make a distinction between relevant and irrelevant information.
Moreover, little literature is available regarding the effect of auditor expertise on a risk assessment during the audit. Therefore, this study considers the literature on the effect of framing of the results by the management with regard to auditor’s decisions. In a broader picture, the unaudited financial statements are a perception of the results of the management. Some research has examined the framing of results by management and the effect that expertise has on the identification of material misstatements regarding fraud and errors. In a qualitative set up, Johnson, Jamal, and Berryman (1991) found that auditors with more industry expertise are better at identifying potential fraud through framing than those without industry expertise regardless of experience, as they have more knowledge of specific cues to look for. Furthermore, they found that auditors with more experience are better at spotting an error than novices. Hence, based on these two findings it is concluded that auditors with more expertise are better at identifying misstatements due to fraud and errors when management has framed the results. Narcissistic people tend to have a unique perception of their own grandiosity and the results they achieve, as explained in Chapter 2.2.1, and could therefore frame the results of their company as well.
12 Based on the research covered in this section, it is presumed that auditor expertise will impact the auditor's perceived risk. At a risk assessment it is almost impossible to develop one right perception of the risk. Based on the arguments presented above I presume that the perceived audit risk at the risk assessment of an audit is significantly different between different between an auditor with low expertise and high expertise.
This leads to the following hypothesis:
Hypothesis 3a: The perception of the audit risk at a risk assessment is significantly different between auditors with a high level of expertise and auditors with a low level of expertise.
Moreover, based on the arguments presented in this section regarding the studies of Krishnan (2003) and DeBoskey and Yang (2012) it is hypothesized that auditors with more expertise are better at identifying audit risks such as earnings management techniques. Besides, auditors with a high level of expertise are better at processing information at a client. Moreover, auditors with auditors with industry expertise are better at identifying potential fraud then auditors without industry expertise. Finally, auditors with more expertise are better at identifying misstatements due to fraud and errors. Therefore, it is concluded that auditors with more expertise will be better at identifying audit risks compared to auditors with less expertise.
In Chapter 2.3 it is hypothesized that a CEO with highly narcissistic tendencies should lead to a higher perceived audit risk by the auditor at an risk assessment compared to a CEO with low levels of narcissistic tendencies. In this study it is presumed that auditors with more expertise are better at identifying the increased audit risk due to a highly narcissistic CEO. This leads to the following hypothesis:
Hypothesis 3b: A higher level of auditor expertise will lead to a higher perceived audit risk by an auditor given a higher level of narcissism of the CEO.
2.5.2 Auditor Narcissism
The concept of narcissism was discussed earlier in Chapter 4. Unlike CEO narcissism, little research in the accounting literature has examined auditor narcissism. Banimahd, Dilami, and Javanmard (2013) conducted an exploratory study on the trait of narcissism among auditors in Iranian accounting firms. They observed a significant difference between different job grades, with more narcissistic tendencies evident among younger auditors. Akers, Giacomino, and Weber (2014) also found a decline in narcissistic traits in the accounting profession from the job grade of associate to manager. In contrast, however, Banimahd et al. (2013) found an increase of narcissism in the partner job grade.
Narcissistic people overestimate their IQ (Gabriel, Critelli, & Ee, 1994) and their own contribution to a collaboration relative to group member ratings and ratings made by independent observers (Gosling, John, Craik, & Robins, 1998). What is more important is that narcissists have distorted judgements of their own capabilities and risk-based decision-making (Twenge et al. 2008). The constant need for admiration rooted in narcissism may drive the narcissist to take excessive risks (Johnson et al., 2012). This distorted judgement has been confirmed by research by Campbell, Goodie, and Foster (2004).
Research is in agreement that auditors in general have an overconfidence bias with regard to the estimation of their colleagues’ technical knowledge (Han, Jamal and Tan (2011). However, it is important that judgements with regard to colleagues be objective and without bias to ensure audit
13 quality (Tan and Jamal, 2001). It is thought that a distorted judgement of capabilities from narcissistic auditors leads to larger biases with regard to the estimation of technical knowledge of colleagues. In their study, Campbell et al. (2004) demonstrated how strong this distorted judgement bias can be: even after underperforming in comparison to peers, narcissists indicated that they had outperformed others. Furthermore, Campbell at al. (2004) found that narcissists’ predictions of future performance were based on expectations rather than actual performance. During a risk assessment, these factors could lead to a lower perceived audit risk considering the overestimating of own capabilities, the client's management integrity, and for example the technical knowledge of team members. Besides this overconfidence bias, narcissists have a higher risk appetite (Lakey, Rose, Campbell and Goodie, 2008). It is presumed here that having a higher risk propensity strengthens the effect of the distorted overconfident decision-making of a narcissistic auditor during a risk assessment. This leads to the following hypothesis:
H4: A highly narcissistic auditor will perceive a lower audit risk compared to non-narcissistic auditors during a risk assessment.
The hypotheses presented in this chapter have been visualized in the conceptual model below. The independent, dependent, experimental, and co-variables are operationalized in Chapter 5.
14
3 Methodology
This chapter explains how the experiment in this study was conducted. It first discusses the population, followed by the setup of the experiment. Subsequently, the variables are operationalized. Finally, the chapter explains the control variables and statistical analysis of the data.
3.1 Population
The participants of this experiment were Dutch professionals working at an audit department of an audit firm. It is important to note that in this study auditors were defined as professionals working at an audit department of an audit firm, and auditors which are certified (CPA) were described as certified auditors in this thesis.
The sample consisted of auditors from different organizations spread throughout the Netherlands to avoid a local office bias and to assure that the sample was representative of all auditors in the country. Furthermore, the professionals performed risk assessments during audit engagements. As criteria for inclusion, the auditors had to be at least senior staff (e.g. 2 years of experience), because in practice they also contribute to the risk assessment process. Auditors from the level of senior staffers are actively participating in the planning phase of the audit to assess the audit risk. Furthermore, it is considered as best practice by audit firms to include all members of the audit team in pre audit meeting in the audit planning phase were the risks of the audit are being assessed. Therefore, it was concluded that it is appropriate to include auditors from the level of senior staff in this experiment.
3.2 Research Design
The experiment was developed and conducted by a research team of the University of Groningen (RUG) and the Vrije Universiteit (VU) and four graduate students of the RUG. In the experiment, professional auditors were asked to perform a risk assessment regarding a fictitious company which was presented in the form of a case. Auditors were asked to assess the fraud risk and audit risk of this case in the planning phase of an audit and to provide some personal background information.
This experiment for this study had a two-by-two design. CEO narcissism and audit committee strength were the experimental variables in this study as shown below in Table 1.
15 After reading the case, participants were asked to complete six lists of questions:
1. Audit and fraud risk assessment 2. Questions regarding CEO narcissism 3. Questions regarding AC strength 4. Risk propensity scale
5. Narcissistic Personality Inventory 16 questionnaire 6. Personal background
Questionnaires regarding the risk assessment and the CEO narcissism (1,2) and the personal background (6) were in line with Johnson et al. (2012). In the personal background questionnaire, items were included regarding expertise, gender, age, and qualification. The questions regarding AC strength (3) were developed by a research team of the RUG and VU. The question sets regarding CEO narcissism and AC strength were included to analyze whether participants observed the high/low narcissism treatment and the strong/weak AC.
The questions regarding risk propensity (4) were based on research by Kneckel et al. (2010).
Finally, for the questions regarding auditor narcissism, a short version of the Narcissistic Personality Inventory (NPI) was used.
Johnson et al.’s (2012) experiment outline was used to improve the internal and external validity. Johnson et al. (2012) used CEO narcissism and opportunity as experimental variables in their research. In contrast, the present study used a case with a CEO instead of a regional director. In this case, all treatments included a high motive and low opportunity for the management to manipulate the financials.
The original case and its manipulations are available upon request at the supervisor of this present research.The variables are operationalized later in this chapter. The experiment was conducted in Dutch.
3.3 Administration
Qualtrics.com software was used to ensure a double-blind experiment. Qualtrics randomly assigned participants, thus participants had a 25% chance of receiving a treatment. The participants received a link in their mailbox to start the risk assessment case in the qualtrics.com portal. As in real risk assessments, it was ensured that the auditors had to rely on their impressions of the company. Hence, it was not possible for participants to look back at the case once the questions started. After completion of the case, it was confirmed that the respondent him- or herself had actually performed the risk assessment. This is assured by asking respondents personally upfront to participate in this study. If they are able to participate a unique link was sent by mail so they could participate in the experiment. Without an unique hyperlink from the invite it was impossible to join this experiment. Finally, an email to thank the respondents was sent after the experiment to confirm that the respondent actually participated in the study themselves. This set up ensured the integrity of the data. To prevent sharing of information between subjects, the case was only open for 6 days, including a weekend.
3.4 Internal and External Validation
To assure that the participants perceived the treatments correctly two questionnaires were included in the experiment after the risk assessment to measure whether the manipulations of CEO narcissism and AC strength were correctly observed by participants. Additionally, three control questions were asked in the CEO manipulation questionnaire with regards to CEO likability, CEO successfulness and CEO competence. This analysis has been included to assess
16 if these factors could have had substantial impact on the risk assessment. These manipulation questionnaires were included after the risk assessment questionnaire to prevent that these questions could have led to a biased risk assessment. Besides, an analysis was made to investigate if the population treatments were homogeneous.
Furthermore, the internal validity of this study was increased by the using the experiment set up of Johnson et al. (2012) and already validated questionnaires. This led to an increased internal validity of the study because these instruments have been used and tested before in the academic literature. The setup was initially reviewed by an independent professor of the VU. Furthermore, a pilot of the experiment was conducted with researchers of the VU and experienced audit professionals. After the pilot, the participants were extensively interviewed to determine whether the experiment measured what it should measure and to ensure that the case had not been misinterpreted by the participants. As previously mentioned, the experiment used a double-blind set up; both the research group and the participants were unaware of the anticipated results. To ensure the generalizability of the study to all auditors in the Netherlands and to avoid local biases, the sample consisted of auditors who worked in different regions of the Netherlands, and included auditors from all Big Four firms in different positions. To increase external validity and reliability of the experiment, the researchers ensured that all participants worked as auditors at an audit firm and had experience conducting risk assessments for actual clients. To make sure that the auditors based their judgement on impressions, just as in real risk assessments, they could not look back at the case once they had started the questionnaires. Finally, to prevent the Hawthorne effect, participants were told that they had to fill in a questionnaire on auditors’ decision-making; thus, they were unaware of the different treatments. The ethical research committee of the RUG had no objections to this study.
3.5 Operationalization of Variables
This section operationalizes the experimental variables and co-variables of this study. This is done for the following variables: CEO narcissism, AC strength, perceived audit risk, and auditor expertise and auditor narcissism.
3.5.1 CEO Narcissism
The narcissism manipulation was largely based on Johnson et al.’s (2012) study. The only difference was that the present study also included a board picture manipulation. Johnson et al. (2012) used several items from Rashkin and Hall’s (1979) NPI. They emphasized the constructs of grandiosity and exploitativeness/entitlement of narcissism.
To increase the internal validity of the present study, the same operational constructs were included as those in Johnson et al.’s (2012) work. A fictitious dialogue with the auditor (participant) and the CEO of the company was included to express the narcissism constructs (high/low) in the experiment. In the high narcissism treatment, the narcissistic traits of the CEO were expressed as follows: grandiosity (“I make sure things get done on time, and let me tell you, it’s not always
easy”), entitlement (“I’ve got my hands full keeping this place running – I don’t͛ have time to do their work for them” and “I worked like a dog to get the change orders and CIP in shape for closing - I must have put 80 hours in that week”), and exploitativeness (“I had a meltdown and told my
staff that they had better follow my procedures to the letter from now on or heads would roll”). In
line with the Johnson et al. experiment, to operationalize the high narcissism treatment, more first-person references (“I”) were included (Chatterjee and Hambrick 2007, 2011; Raskin and Shaw 1988), along with the tendency for the CEO to react with anger to an ego threat (Rhodewalt and Morf 1998; Stucke and Sporer 2002).
Finally, a board picture was included to further emphasize the grandiosity element of the NPI. In prior research, Olsen et al. (2013) used this method to measure narcissistic behavior of managers.
17 In the high CEO narcissism treatment, a picture of the board team was included where the CEO stood in the foreground, while other board members were placed behind him in the background. In contrast, in the low CEO narcissism treatment, the CEO was placed between the other board members without a predominant position. The cases that have been used for the four different treatments and the questionnaires are available upon request at the supervisor of this study.
3.5.2 AC Strength
AC strength was operationalized with the constructs of AC activity (number of meetings), AC experience, and AC independence. This elements have been discussed in the Chapter 2.4. The development of hypothesis 2 explained why these proxies were included to differentiate between a weak and a strong AC.
In the strong AC treatment, the AC had six yearly meetings, in contrast to semi-annual meetings in the weak AC treatment. AC experience was manipulated by an ex-Big Four audit partner in the strong AC treatment in comparison to an experienced purchase manager in the weak AC treatment and an AC member with a background in the banking industry in the strong treatment, versus a retired HR consultant in the weak treatment. For AC independence, the proxy of social ties of the AC with the CEO was used. This was operationalized by including a golf friend and a friend active in other social networks with the CEO in the weak AC treatment, and an ex-Big Four audit partner in the strong AC treatment.
3.5.3 Perceived Audit Risk
To measure the dependent variable, Johnson et al.’s (2012) questionnaire was adapted. Johnson et.al used a 7-point Likert scale to measure fraud risk, from very low to very high. In the present study, two questions were asked: one regarding the fraud risk and one regarding the audit risk. Based on the literature in Chapter 2 it was hypothesized that audit risk is different from fraud risk. In this study with a Cronbach’s alpha test it is measured if the perceived fraud risk is different from the perceived audit risk. This analysis was done to ensure that the perceived audit risk variable measured what it was intended to do and that it does not measure the fraud risk.
3.5.4 Auditor Expertise
As shown in the literature review, Beck and Solomon (1989) argue that experience is unlikely to be a surrogate for expertise in many public accounting tasks. Therefore, more task-specific elements were chosen in this study instead of a general experience element such as experience in years. Ashton (1991) shows that auditing expertise should not be assumed as a general proxy that extends to all auditing tasks and all industries. Hence, in this study, the variable expertise was defined as the combination of personal expertise factors relevant to the experiment: qualification, misstatements come across during an audit, and number of clients audited in the industry. Moreover, this expertise variable was defined based on the proxies available in of Johnson et al.’s (2012) questionnaires, which were also used in this experiment.
An important proxy used for “general” expertise in this research was qualification. This proxy shows that from the perception of the Dutch association of auditors, the NBA, an auditor has sufficient expertise to sign off on an audit report. To keep this title, yearly education must be followed and since 2017, annual tests (the “kennistoets”) are held by the NBA for Dutch CPAs. Hence, qualification was included in the operationalization of auditor expertise as a general proxy for expertise.
In addition, as a proxy for more specific expertise, this study used the number of material fraud errors the auditor had encountered in the past; this is in line with Ashton’s (1991) work. The study of Ashton suggests that even general auditors have limited direct experience with financial statement errors and that auditors only know the most frequent causes of an error.
18 A potential problem with this proxy was that auditors did not know the exact number of material errors they had encountered in their career, which potentially led to a measurement that had either a high probability of an estimation bias or a non-response bias. However, it was presumed that auditors were better at recapitulating the material errors they had come across due to fraud, and that this measurement was therefore less subject to these biases.
Finally, a measure for industry expertise was included as a proxy for auditor expertise. In the literature review, in was shown that clients of industry expert auditors report lower accruals and therefore higher quality. Moreover, clients audited by an industry expert are more likely to receive a modified audit opinion compared to those audited by non-specialists (Chi and Chin, 2011). This suggests that industry specialists more often find and report material errors in financial statements.
Industry expertise was operationalized as the frequency with which auditors audited clients in this sector. A compromise was made here, by including a more experience-related component instead of a pure expertise component in the expertise variable. This is because industry expertise is difficult to measure in an experiment. Furthermore, it is expected that having performed more risk assessments in the related industry should lead to a greater knowledge of the potential risks that could lead to increased audit risk.
The three elements based on which expertise was measured – qualification, number errors due to fraud encountered at an audit, and industry expertise – were equally weighted. An index, auditor expertise has been created as shown in the Table 1 below.
3.5.5 Auditor Narcissism
In the literature, there are two widely used measures for narcissism. These are the NPI by Rashkin and Hall (1979), and the Millon Clinical Multiaxial Inventory (MCMI). In line with Akers et al.’s (2014) study, which investigated narcissism in accounting firms, the present study measured the narcissism of auditors using the NPI. Moreover, the MCMI was used to measure clinical narcissism instead of narcissism observed in the general population.
The classical NPI consists of 40 questions. For the present study, however, this question set would have been too large and would have led to a larger non-response rate of the participants, as it would have taken a more substantial time investment. Therefore, Ames, Rose, and Anderson’s (2005) NPI-16 was used. These authors suggest that “the NPI-16 is a valid way to capture the construct narcissism in situations in which use of a more extensive measure would be impractical.”
19
3.6 Control Variables
In this study there was controlled for auditor gender and auditor general experience in accountancy. Gender was controlled for because research shows that males and females process information differently during the analytical procedures of an audit (O’Donnel & Johnson, 2001). In this vein, Niskanen et al. (2011) suggest that female auditors are more conservative than their male colleagues. This conservativeness could impact risk assessments. Besides, there has been controlled for general audit experience in accountancy to investigate further the effects of auditor expertise and auditor general experience. Due to multicollinearity issues with age and general auditor experience accountancy, there has only been controlled for general audit experience in the regression analysis.
To summarize the variables an overview of the variables has been included below.
3.7 Data analysis method
The ordinary least square regression analysis was used in this study. This method had the preference over the ANCOVA analysis, because the regression analysis had a better fit with this study. ANCOVA is used for analysis to provide insight between different groups of data, where regression gives estimates of the relationships among variables.
The linear regression analysis provides much more flexibility and plotting options in comparison to the ANCOVA analysis, moreover it takes more effort to test if the data fits the assumptions for an ANCOVA analysis in comparison to a linear regression analysis given the setup of this experimental study (Field, 2009). Therefore, it is concluded that the linear regression analysis is the most suitable to this study.
20
4 Data Analysis
The effects of the experimental conditions for CEO narcissism and AC strength were modeled in a 2 (High vs Low Narcissism) x 2 (Weak vs Strong AC) experimental design, as shown in Table 2 in Chapter 3.5.
Table 4 shows a summary of the descriptive statistics of the experiment and the participants. These descriptive statistics are included to provide an overview of the population and experiment set up. Appendix A contains more detailed descriptive statistics of the variables used in this experiment.
4.1 Manipulation Check Tests of CEO Narcissism and AC strength
This subsection contains the manipulation check analyses performed for CEO narcissism and AC strength. Both CEO narcissism and AC strength have been manipulated across the treatments. The first step in the analysis involved assessing whether the respondents observed the manipulations correctly. This was done to demonstrate the effectiveness of the manipulations. It is concluded that the participants successfully interpreted the experimental treatments.
The CEO narcissism manipulation questionnaire consisted of nine questions. Six of these apply to the manipulation check for CEO narcissism. The other three questions were control questions to assess if CEO likability, CEO successfulness and CEO competence had a potential influence on the risk assessment.
21 An internal consistency test for the CEO narcissism manipulation using the Cronbach's alpha test indicated a “questionable” to almost “acceptable” quantitative internal consistency (0.694) between the different CEO narcissism manipulation questions. Deleting questions would only improve this score by 0.015. The overall internal consistency when taking into consideration the qualitative consistency is sufficient because certain personality characteristics of narcissistic behaviour, such as grandiosity, were used as to measure narcissism and narcissism was not directly measured. Hence, the questions were grouped as one variable for the CEO narcissism manipulation test. An independent samples test showed that the variance in the questions with regards to the CEO narcissism manipulations is significant between the CEO narcissism treatments. Additionally, the results in Table 2 show that the respondents of the high CEO narcissism treatment indicated a higher perceived audit risk than the low CEO narcissism treatment. Therefore, the CEO narcissism manipulation was correctly identified by the participants in this experiment.
The same analysis was performed for the AC strength manipulation check. The internal consistency of the four questions for the manipulation check of AC strength is “good” (0.882). The internal reliability could not be further improved by deleting questions. Based on the Cronbach's alpha score, the AC strength manipulation questions were grouped together as one variable. The results indicate that the difference in variance between the treatments for the AC strength manipulation questions is significant. Moreover, Table 2 indicates that respondents in the strong AC strength treatment scored a lower perceived audit risk on average compared to the Weak AC strength treatment. Therefore, the AC strength manipulation was successful in this experiment.
4.2 Data
4.2.1 Missing Data
For 2 of the 125 participants, the age is unknown. This missing data problem has been solved by mean imputation. Additionally, two auditors did not provide their experience but only their position. For these two cases, the general required experience for that position was included. This imputation of the expected required general experience prevented an overstatement of the experience which could have led to a bias in the regression analysis results of this study.
4.2.1 Interaction Effects
The interaction effect between CEO Narcissism and AC Strength has been calculated by using dummy variables. For the interaction effect between CEO narcissism and auditor expertise, the dummy variable of CEO Narcissism has been multiplied by the standardized value of Auditor Expertise.
4.3 Other Analyses
In addition to the manipulation checks for the CEO narcissism and the AC strength treatment, the control questions in the CEO narcissism manipulation questionnaire were analyzed. This analysis tested whether other perceptions of the CEO personality could have led to a bias in the perceived audit risk for this study. An independent samples test has been performed on the CEO narcissism control questions regarding competence, likability and successfulness of the CEO. As shown in Table 7 in Appendix A, the variance is significantly (two tailed p-value <0.00) only for CEO likability between the different treatments of CEO narcissism (high/low).
Moreover, the results show that in the high CEO narcissism treatment the CEO was perceived as less likable (3.08 on a scale from 1-7) in comparison to the low CEO narcissism treatment (4.24 on a scale from 1-7). Therefore, based on the significant effect in variance, there was controlled
22 for this CEO likability in the regression analysis in Table 6 to assure that the CEO likability was not the predictor for the perceived audit risk by the respondents between different treatments groups.
Next, additional analyses were performed to confirm the lack of population biases among treatments. The male to female ratio and average experience in accountancy is similar between the different treatments; hence no problems with the spread of gender and age of respondents between different treatments was identified that could lead to a bias in our linear regression analysis regarding the experimental conditions CEO narcissism and AC strength. Moreover, the risk propensity of the four treatment groups was similar.
Additionally, a non-response analysis was conducted, and the results show a similarity between the male to female ratio of the non-respondents and participants of the experiment. This shows there was no non-response bias regarding gender of the non-respondents. Furthermore, the results show a higher amount of non-response with respect to experience and more senior auditors. This is expected as more senior auditors may have limited availability. Enough senior auditors participated in the experiment to prevent an experience bias due to non-response. A Cronbach’s alpha consistency analysis was performed between the participant’s perceived fraud risk and the perceived audit risk. The results indicate that audit risk and fraud risk have a questionable quantitatively consistency (0.616). This suggests that the variables fraud risk and audit risk are correlated but do not measure the same construct. This is in line with the argumentation in the literature review in Chapter 2. These results of the internal consistency check indicated that the respondents made a distinction between fraud risk and audit risk.
Finally, it is noted that the data did not violate the assumptions for linear regression: homoscedasticity, linearity, or normality of the residuals. Outliers in the data did not have a negative impact on the regression analysis. Multicollinearity will be discussed in the next Subsection.
Multicollinearity analysis
The multicollinearity of the variables was analyzed pre-test and post-test. Table 5, at the next page, shows the correlations matrix. A high correlation (0.958) was shown between Age and General Experience in Accountancy. Including both in the regression analysis would lead to multicollinearity issues. Therefore, for General Experience in Accountancy was controlled in this study, but not for Age since both variables appear to measure the same construct.
In addition, a high correlation was shown between General Experience in Accountancy and Auditor Expertise. Similarly, a high correlation was shown between the Audit Expertise and the interaction effect of CEO Narcissism and Auditor Expertise. However, the variance inflation factor multicollinearity diagnostics post-test shows that there are no multicollinearity problems between the variables (The VIF values of the variables are below 3.1).
23
4.4 Control Variables Analysis
In this study, there has been controlled for the variables CEO Likability, Gender and Auditor General Experience in Accountancy. Table 6 shows that when only including the control variables (model 1), the variables CEO Likability (two tailed p-value < 0.05), and General Auditor Experience in Accountancy (two tailed p-value < 0.01) are significant. However, after including all variables of the conceptual model of Chapter 2 and the control variables in the regression model (model 4 in Table 6), CEO Likability is insignificant (two tailed) and Auditor General Experience in Accountancy is significant (two tailed p-value < 0.01). This suggests that CEO Likability does not have a significant influence on the perceived audit risk of the respondents in this study. For auditor gender, the results show an insignificant (two tailed) effect. This means that the gender of the auditor did not have an impact on the perceived audit risk. Furthermore, the control variable Auditor General Experience in Accountancy has a significant effect (two tailed p-value < 0.1) when including all the control variables in the final research model (model 3 in Table 6). Therefore, the effect of this control variable is significant and is included in the final research model (model 4 in Table 6). Multicollinearity between the variables Auditor Expertise and Auditor General Experience in accountancy was sufficiently low in this regression model. This multicollinearity was analyzed using variance inflation factors. These variance inflation factors are below 3.1 in the regression models included in Table 6.
24
4.5 AC Strength and CEO Narcissism Effects (Hypotheses 1, 2a and 2b)
The hypotheses were tested with the regression Model 4 shown in Table 6. This model contains all the variables included in the conceptual model described in Chapter 2 and the significant control variable General Auditor Experience in Accountancy.
The effect of the CEO narcissism treatment was insignificant (two tailed) in the research model when including the AC Strength and CEO Narcissism variable, their interaction, the co-variables Auditor Expertise and Narcissism, and the interaction effect between CEO Narcissism and Auditor Expertise. Therefore, these results do not support H1.
Furthermore, it is hypothesized that the AC strength treatment affected the perceived audit risk of the respondent. Based on the results of the final research model with the significant control
